CN107992368A - Method for interchanging data and system between a kind of multi-process - Google Patents
Method for interchanging data and system between a kind of multi-process Download PDFInfo
- Publication number
- CN107992368A CN107992368A CN201711132105.2A CN201711132105A CN107992368A CN 107992368 A CN107992368 A CN 107992368A CN 201711132105 A CN201711132105 A CN 201711132105A CN 107992368 A CN107992368 A CN 107992368A
- Authority
- CN
- China
- Prior art keywords
- data
- node space
- space
- node
- write
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/54—Interprogram communication
- G06F9/543—User-generated data transfer, e.g. clipboards, dynamic data exchange [DDE], object linking and embedding [OLE]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/52—Program synchronisation; Mutual exclusion, e.g. by means of semaphores
- G06F9/524—Deadlock detection or avoidance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/52—Program synchronisation; Mutual exclusion, e.g. by means of semaphores
- G06F9/526—Mutual exclusion algorithms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/54—Interprogram communication
- G06F9/546—Message passing systems or structures, e.g. queues
Abstract
The present invention relates to the method for interchanging data and system between a kind of multi-process, include the use of shared drive and domain name socket mixed architecture exchanges data, and shared drive is divided into N number of node space, data sharing is carried out, and the data in shared drive are synchronized using domain name socket.Since each process uses mutual exclusion lock access identities position before accessed node space, avoid rdma read and write the conflict between memory.The collocation of the flag and mutual exclusion lock of node space uses, so that mutual exclusion between data production process and multiple data consuming process, and can be accessed at the same time between multiple data consuming process with a piece of memory so that data consuming process will not cause because of locking and caused by hydraulic performance decline.
Description
Technical field
The present invention relates to information technology field, method for interchanging data and system between more particularly to a kind of multi-process.
Background technology
Data Interchange Technology between the multi-process being present invention is primarily concerned with point in host, is especially suitable between multi-process
Big data exchanges.
Common Data Exchange includes the use of pipeline, message format communication, shared drive, socket communication
(socket), domain name socket (domainsocket etc.).Data exchange amount between process can use the above when smaller
Mode is completed, but when needing to exchange substantial amounts of data between process, these modes can be because processing data bandwidth be high, it is more to take
Additional CPU time and the requirement that cannot meet real-time, while the sending action of domain name socket is by the traps of operating system
Triggering, and the resource of traps is limited, also limit the treatable maximum bandwidth of which institute.
Shared drive is then that a part of region is marked off in the memory of operating system as two or more process in system
The memory being used in conjunction with.This mode make between process data exchange need not be replicated, therefore be adapted to large-scale data
Exchange.But the data exchange based on shared drive needs to carry out complicated data simultaneously operating, and the degree of coupling is high between causing process,
Any one process goes wrong, other all processes will appear from blocking.Such as the side that data production process passes through shared drive
Formula provides data to multiple data consuming process, if one of data consuming process goes wrong, cause cannot be in shared
It is insufficient it will cause shared memory space to take data in depositing away in time, ultimately results in data production process and is also stopped, so that
Whole processing system is caused to be stopped.
For avoid shared drive indivedual process failures in cause all processes deadlock situation, the present invention propose make
Realize that the large-scale data in host between multi-process exchanges with shared drive with mode that socket combines.Specifically, to carry
The data exchange efficiency of high socket, the synchronizing information of data is transmitted using domain name socket.Moreover, in domain name socket only
Transmit the offset of shared drive, can in once transmitting synchronous multiple data slots, can be greatly decreased to operating system
The use of traps resource, reduces demand of the interactive information to bandwidth, improves the efficiency of data exchange between multi-process.This method
Not only the characteristics of having make use of shared drive data exchange bandwidth high, but also eliminated using domain name socket between process to internal storage access
Coupling caused by deadlock.
The content of the invention
The defects of it is an object of the invention to overcome above-mentioned technology, there is provided a kind of based on shared drive and domain name socket
Large-scale data exchange method and system between multi-process in host.
Specifically, the invention discloses the method for interchanging data between a kind of multi-process, including：
Data consuming process starting step, starts multiple data consuming process, and empty to operating system application shared drive
Between；
Shared memory space partiting step, starts a data production process, the data production process and the data consumption
Process defines Data Management Unit, and the form by the shared memory space according to the Data Management Unit using identical rule
N number of node space is divided into, wherein N is positive integer；
First node space selecting step, chooses a certain node space as node space to be written；
Data write step, the data production process access the flag of the node space to be written, if the node to be written is empty
Between flag be idle condition, then be set in write state, while the data production process is write to the node space to be written
Enter data, by home position be idle condition after the completion of write-in, if the flag of the node space to be written is in read states, put
The node space to be written is abandoned, chooses next node space as the node space to be written；
The data write step is called in synchrodata tuple forwarding step, repetition, the number which is produced
According to writing successively in N number of node space, and num node space is often write just by domain name socket by synchrodata tuple
The data consuming process is sent to, wherein num is the positive integer for being less than N more than 1；
Section point space selecting step, the data consuming process obtain enabled node by parsing the domain name socket
Space, and according to the enabled node space, a certain node space is chosen as the node space that continues；
Data reading step, the data consuming process access the flag of the node space that continues, if this continues, node is empty
Between flag for idle condition or in read states, then be set in read states, while the data consuming process reads and locates
Data in the node space that continues are managed, by home position are idle condition after the completion of processing, the mark of node space if this continues
Position is in write state, then abandons the node space that continues, and according to the enabled node space, selection next node space, which is used as, treats
Read node space；
Repeat step, the data reading step is called in repetition, until all data consuming process complete the reading of data
Take and handle.
Method for interchanging data between the multi-process, wherein the data write step further include, the data production process to this
During node space write-in data to be written, if it is former to write direct new data cover with data for the node space to be written
Some data.
Method for interchanging data between the multi-process, wherein the data production process access node sky to be written using mutual exclusion lock
Between flag；And the data consuming process accesses the flag of the node space that continues using mutual exclusion lock.
The circulation of method for interchanging data between the multi-process, wherein the data production process is write in N number of node space successively
Enter data and sent N number of nodal information to data consuming process by domain name socket.
The invention also discloses the data exchange system between a kind of multi-process, including：
Data consuming process starting module, for starting multiple data consuming process, and in sharing to operating system application
Deposit space；
Shared memory space division module, for starting a data production process, the data production process and the data
Consuming process defines Data Management Unit using identical rule, and by the shared memory space according to the Data Management Unit
Form is divided into N number of node space, and wherein N is positive integer；
Module is chosen in first node space, for choosing a certain node space as node space to be written；
Data write. module, for making the data production process access the flag of the node space to be written, if this is to be written
The flag of node space is idle condition, then is set in write state, while the data production process is to the node to be written
Space writes data, by home position be idle condition after the completion of write-in, if the flag of the node space to be written is in reading shape
State, then abandon the node space to be written, chooses next node space as the node space to be written；
Synchrodata tuple sending module, for repeating to call the Data write. module, which is produced
Data write successively in N number of node space, and often write num node space just by domain name socket by synchrodata
Tuple is sent to the data consuming process, and wherein num is the positive integer for being less than N more than 1；
Module is chosen in section point space, for making the data consuming process parse the domain name socket, obtains available section
The space of points, and according to the enabled node space, a certain node space is chosen as the node space that continues；
Data read module, for making the data consuming process access the flag of the node space that continues, if this continues
The flag of node space is then set in read states for idle condition or in read states, while the data consuming process is read
Take and handle data in the node space that continues, by home position be idle condition after the completion of processing, the node space if this continues
Flag be in write state, then abandon the node space that continues, according to the enabled node space, choose next node space
As the node space that continues；
Replicated blocks, for repeating to call the data read module, until all data consuming process complete data
Reading and processing.
Data exchange system between the multi-process, the wherein Data write. module further include, the data production process to this
During node space write-in data to be written, if it is former to write direct new data cover with data for the node space to be written
Some data.
Data exchange system between the multi-process, wherein the data production process access node sky to be written using mutual exclusion lock
Between flag；And the data consuming process accesses the flag of the node space that continues using mutual exclusion lock.
The circulation of method for interchanging data between the multi-process, wherein the data production process is write in N number of node space successively
Enter data and sent N number of nodal information to data consuming process by domain name socket.
Compared with prior art, the advantage of the invention is that：
First, data between multi-process are improved relative to directly using domain name socket to exchange data using shared drive
The bandwidth of exchange.Shared drive is divided into N number of node space, improves the service efficiency of shared drive between multi-process.Secondly,
Favorable expandability.Communicate due to the use of domain name socket so that large-scale data exchange can between multiple processes in order into
Go and be not limited solely between two processes.Meanwhile data production process employs the strategy of directly covering legacy data, even in number
In the case of cannot taking data away according to consuming process, data production process is appointed so can write data in the shared drive of annular
Keep normal work.Data consuming process can obtain data after recovering normal and work on immediately.Finally, due to each process
Mutual exclusion lock access identities position is all used before accessed node space, rdma read is avoided and writes the conflict between memory.Node
The collocation of the flag and mutual exclusion lock in space uses, as long as being related to flag rewriting, will use mutual exclusion, first obtain, then discharges,
So that mutual exclusion between data production process and multiple data consuming process, and can be accessed at the same time between multiple data consuming process
With a piece of memory, though such design increase more data consuming process will not cause because locking and caused by under performance
Drop.
Brief description of the drawings
Fig. 1 is data production process flow chart；
Fig. 2 is data consuming process flow
Fig. 3 is the implication tabular drawing of each of flag；
Fig. 4 is annular shared drive schematic diagram when applying for N number of isometric continuous shared memory space.
Embodiment
The invention discloses the method for interchanging data between a kind of multi-process, including：
Data consuming process starting step, starts multiple data consuming process, and empty to operating system application shared drive
Between；
Shared memory space partiting step, starts a data production process, the data production process and the data consumption
Process defines Data Management Unit, and the form by the shared memory space according to the Data Management Unit using identical rule
N number of node space is divided into, wherein N is positive integer；
First node space selecting step, chooses a certain node space as node space to be written；
Data write step, the data production process access the flag of the node space to be written, if the node to be written is empty
Between flag be idle condition, then be set in write state, while the data production process is write to the node space to be written
Enter data, by home position be idle condition after the completion of write-in, if the flag of the node space to be written is in read states, put
The node space to be written is abandoned, chooses next node space as the node space to be written；
The data write step is called in synchrodata tuple forwarding step, repetition, the number which is produced
According to writing successively in N number of node space, and num node space is often write just by domain name socket by synchrodata tuple
The data consuming process is sent to, wherein num is the positive integer for being less than N more than 1；
Section point space selecting step, the data consuming process obtain enabled node by parsing the domain name socket
Space, and according to the enabled node space, a certain node space is chosen as the node space that continues；
Data reading step, the data consuming process access the flag of the node space that continues, if this continues, node is empty
Between flag for idle condition or in read states, then be set in read states, while the data consuming process reads and locates
Data in the node space that continues are managed, by home position are idle condition after the completion of processing, the mark of node space if this continues
Position is in write state, then abandons the node space that continues, and according to the enabled node space, selection next node space, which is used as, treats
Read node space；
Repeat step, the data reading step is called in repetition, until all data consuming process complete the reading of data
Take and handle.
Method for interchanging data between the multi-process, wherein the data write step further include, the data production process to this
During node space write-in data to be written, if it is former to write direct new data cover with data for the node space to be written
Some data.
Method for interchanging data between the multi-process, wherein the data production process access node sky to be written using mutual exclusion lock
Between flag；And the data consuming process accesses the flag of the node space that continues using mutual exclusion lock.
The circulation of method for interchanging data between the multi-process, wherein the data production process is write in N number of node space successively
Enter data and sent N number of nodal information to data consuming process by domain name socket.
Specifically, the present invention is realized by following technical solution：
A. start multiple data consuming process first, be total to from one in multiple data consuming process to operating system application
Memory headroom is enjoyed, all data consuming process begin preparing for receiving domain name socket after the completion of application.One number of subsequent start-up
According to production process.
B. data production and consumption process defines the structure of Data Management Unit using identical rule, is wrapped in structure
Array (memory headrooms of actual storage data) containing mutual exclusion lock, flag flag and actual storage data.The reality of array
Size is Block (Block can be sized according to practical application request, such as 5MB, 100MB etc.).By shared memory space
Form according to Data Management Unit is divided into the identical node space of N number of size, and N is positive integer.This N number of node space is from 0
Start to finish sequence number in order, the numbering of last node space is N-1.
C. data production process writes data into a node space.Before a node space is write data into, make
The flag of the node space is accessed with mutual exclusion lock, if just in reading evidence, i.e., flag is displayed without data consuming process
It is 0 to read flag, then being that the current node space of representative is in write state by home position writes data mode, discharges mutual exclusion
Lock, then starts to write data into node space.The same mutual exclusion lock that obtains is by flag after data production process writes data
It is set to idle condition and then discharges mutual exclusion lock.The data of generation are write in this N number of node space successively along these lines,
Num node space is often write just by domain name socket by synchrodata tuple<num,index_1,index_2,…,
index_num>Corresponding data consuming process is sent to, each node space for having been written into data is included in synchrodata tuple
Numbering.Here num is super ginseng, can be sized according to the frequency of application data exchange, and num is just whole less than N more than 1
Number, while it is less than min (MTU/sizeof (int) -1, N), wherein MTU is the maximum biography that the domain name socket that system is set uses
Byte number shared by size (being usually 1500 bytes) wherein sizeof (int) expression integers of defeated unit, uses sizeof
(int) so that calculation is independently of different platforms.If data production process finds have data consuming process reading
The node spaces of data then this node space of abort operation will be write, next node space is changed into, to avoid conflict.
The circulation of data production process writes Data Concurrent successively in this N number of node space and send synchrodata tuple.Such strategy makes
N number of node space to logically create an annular region as shown in Figure 4.If node space has been had been written into data
Hold, then write direct the original content of new data cover.This N number of node space of such circulate operation.
D. after data consuming process receives domain name socket, the content in domain name socket is write to the buffering of oneself
Area.The content of domain name socket is parsed, obtains the numbering of pending each node space.It is first before handling node space data
The flag of the node space first is accessed using mutual exclusion lock, is shown if flag if data production process is currently written into data
Abandon handling the content of the node space, directly handle the pending node space such as next.If flag is shown as empty
It is not busy or in read states, then the data in the node space are read, and handle.Change flag using mutual exclusion lock after being disposed
State and discharge mutual exclusion lock.
To allow features described above of the invention and effect to illustrate more clearly understandable, special embodiment below, and coordinate
Bright book attached drawing is described in detail below.
Identical node space structure is defined at the same time in data production system A and data consumption system B, C：
Wherein, flag node datas read and write flag, are claimed as char types in example, each process uses reading and writing mark
Know each 1, can at most support the data exchange between 4 processes, can be as needed, the data type of the field is changed, with branch
Hold more multi-process；Pkt_len is used for recording the size of data of this node space actual storage, and pkt_data [MTU_size] is used
Carry out actual storage valid data, pmutex is mutual exclusion lock.
It is now assumed that the scene of a data exchange, including a data production process A and two data consuming process are
System B and C.
To the shared drive of operating system application sizeof (_ pkt_node) * N bytes, (N is node to data consumption system B
The number in space), obtain the ID and initial address of shared drive.Other systems obtain memory ID and rise in the same way
Beginning address.In each system, initial address is stored in (void*) shm, passes through sentence struct_pkt_node*node=
(struct_pkt_node*) shm so that node+i can access the node space of serial number i.
The identical structure for being used for domain name socket and sending data is also defined in data production and consumption system：
Wherein pkt_num=NumOfpktPerTime represents that data production system A often writes full NumOfpktPerTime's
Node space just sends a domain name socket.node_index
For storing the sequence number of each node space.Sending domain name socket in view of system A call operations system can make
Into traps, excessively can frequently packet loss be caused to decline system performance if sending domain name socket, it is excessively slow to cause data again
It is not in time for processing also and new data is just write by data production system A erasings, therefore, by testing repeatedly, it is proposed that will
NumOfpktPerTime is arranged to 31.31 be not a value fixed all the time, can be adjusted according to actual conditions.To avoid counting
According to few abnormal conditions, timeout mechanism can be introduced, after overtime, timely data volume is several at that time by sending to reach preset value
According to.
Since data generation system A needs to send domain name socket to data consumption system B, C, therefore needed just in A
Beginningization domain name socket simultaneously binds B, the domain name socket addresses of C.
As shown in Figure 1, after completing shared drive and the initialization of domain name socket, instantiation _ pktnode_msg in A systems
Structure is msg.Data production system A determines whether that data need to provide and arrives data consumption system, if do not continued
Treat, if then starting to write data into each node space.Need to lock before data write-in node space and access the space
Flag, read flag according to Fig. 3 and judge the space whether B, C reading, and has system to show in meter reading unavailable, otherwise represent can
With, at the same will be available whether reflect a Boolean type variable.According to Fig. 3 by corresponding flag positions if the space can use
1. then flag is unlocked.The code of this process is as follows：
Represent that this node space is unavailable if variable deal_or_not is false, abandon straight to current spatial
Connect and next node space is operated.If true, then corresponding content is write in node space, and will actually write
In the pkt_len of the size of data write-in node space structure entered.At the same time by the sequence number write-in msg's of this node space
In node_index.And then secondary locking operates flag, as Fig. 3 by relevant position 1 and is set to 0.
pthread_mutex_lock(&((node+i)->pmutex))；//lock
(node+i)->Flag＆=~1；
(node+i)->Flag |=1<<1；
pthread_mutex_unlock(&((node+i)->pmutex))；//unlock
If the node space number having been written into is equal to NumOfpktPerTime, structure msg is passed through into domain name set
Word is connect to be sent in corresponding data consumption program.
The course of work of data consumption program B, C are as shown in Figure 2.
Similar with system A, B, C use a positive integer application shared drive identical with A, obtain shared drive ID and rise
Beginning address (void*) shm.Initialize B, the domain name socket of C and the pointer for applying for a struct_pktnode_msg* type
Msg, and itself and char arrays msg_buf are bound together, msg pointers easy to use directly access data.Use header file
Recv functions in sys/socket.h receive domain name socket.If do not receive message to continue to wait, after receiving message
Information write-in msg_buf buffering areas in domain name socket are realized that code is as follows：
Struct_pktnode_msg*msg=(struct_pktnode_msg*) msg_buf；
NumBytes=recv (sockfd, msg_buf, 1024,0)；
Wherein numBytes is the size of data being an actually-received.
Msg- is successively read after receiving data>Node space sequence number in node_index arrays simultaneously is attempted to handle.Read
Need to lock the flag for accessing the space before data in node space, read flag, according to Fig. 3 and judge whether the space has
System is being write, and represents unavailable writing, and reflects whether representing available in reading or idle condition, while node space can use
One Boolean type variable.According to Fig. 3 by corresponding flag positions 1 if the space can use, then flag is unlocked.This process
Code it is as follows：
pthread_mutex_unlock(&((node+msg->node_index[i])->pmutex))；
The processing to this node space is abandoned if deal_or_not is false, reads next node space
Sequence number, deal_or_not are that the data that true is then read in node space are handled.Locked again after the completion of reading to flag into
Row operation.So circulation receives message and handles.
It is below system embodiment corresponding with above method embodiment, this implementation system can be mutual with the above embodiment
Coordinate and implement.The above-mentioned relevant technical details mentioned in mode of applying are still effective in this implementation system, in order to reduce repetition, this
In repeat no more.Correspondingly, the relevant technical details mentioned in this implementation system are also applicable in the above embodiment.
The invention also discloses the data exchange system between a kind of multi-process, including：
Data consuming process starting module, for starting multiple data consuming process, and in sharing to operating system application
Deposit space；
Shared memory space division module, for starting a data production process, the data production process and the data
Consuming process defines Data Management Unit using identical rule, and by the shared memory space according to the Data Management Unit
Form is divided into N number of node space, and wherein N is positive integer；
Module is chosen in first node space, for choosing a certain node space as node space to be written；
Data write. module, for making the data production process access the flag of the node space to be written, if this is to be written
The flag of node space is idle condition, then is set in write state, while the data production process is to the node to be written
Space writes data, by home position be idle condition after the completion of write-in, if the flag of the node space to be written is in reading shape
State, then abandon the node space to be written, chooses next node space as the node space to be written；
Synchrodata tuple sending module, for repeating to call the Data write. module, which is produced
Data write successively in N number of node space, and often write num node space just by domain name socket by synchrodata
Tuple is sent to the data consuming process, and wherein num is the positive integer for being less than N more than 1；
Module is chosen in section point space, for making the data consuming process parse the domain name socket, obtains available section
The space of points, and according to the enabled node space, a certain node space is chosen as the node space that continues；
Data read module, for making the data consuming process access the flag of the node space that continues, if this continues
The flag of node space is then set in read states for idle condition or in read states, while the data consuming process is read
Take and handle data in the node space that continues, by home position be idle condition after the completion of processing, the node space if this continues
Flag be in write state, then abandon the node space that continues, according to the enabled node space, choose next node space
As the node space that continues；
Replicated blocks, for repeating to call the data read module, until all data consuming process complete data
Reading and processing.
Data exchange system between the multi-process, the wherein Data write. module further include, the data production process to this
During node space write-in data to be written, if it is former to write direct new data cover with data for the node space to be written
Some data.
Data exchange system between the multi-process, wherein the data production process access node sky to be written using mutual exclusion lock
Between flag；And the data consuming process accesses the flag of the node space that continues using mutual exclusion lock.
The circulation of method for interchanging data between the multi-process, wherein the data production process is write in N number of node space successively
Enter data and sent N number of nodal information to data consuming process by domain name socket.
Although the present invention is disclosed with above-described embodiment, specific embodiment only to explain the present invention, is not used to limit
The present invention, any those skilled in the art of the present technique, without departing from the spirit and scope of the invention, can make the change and complete of some
It is kind, therefore the scope of the present invention is subject to claims.
Claims (10)
- A kind of 1. method for interchanging data between multi-process, it is characterised in that including：Data consuming process starting step, starts multiple data consuming process, and to operating system application shared memory space；Shared memory space partiting step, starts a data production process, the data production process and the data consuming process Data Management Unit is defined using identical rule, and the shared memory space is divided according to the form of the Data Management Unit For N number of node space, wherein N is positive integer；First node space selecting step, chooses a certain node space as node space to be written；Data write step, the data production process access the flag of the node space to be written, if the node space to be written Flag is idle condition, then is set in write state, while the data production process writes number to the node space to be written According to by home position being idle condition after the completion of write-in, if the flag of the node space to be written is in read states, abandon this Node space to be written, chooses next node space as the node space to be written；Synchrodata tuple forwarding step, repetition call the data write step, by the data production process produce data according to In the secondary N number of node space of write-in, and often write num node space and just sent synchrodata tuple by domain name socket To the data consuming process, wherein num is the positive integer for being less than N more than 1；Section point space selecting step, the data consuming process obtain enabled node space by parsing the domain name socket, And according to the enabled node space, a certain node space is chosen as the node space that continues；Data reading step, the data consuming process access the flag of the node space that continues, the node space if this continues Flag is then set in read states, while the data consuming process reads and handles this for idle condition or in read states Continue data in node space, by home position is idle condition after the completion of processing, if this continues, the flag of node space is In write state, then the node space that continues is abandoned, according to the enabled node space, choose next node space as the section that continues The space of points；Repeat step, repetition call the data reading step, until all data consuming process complete data reading and Processing.
- 2. the method for interchanging data between multi-process as claimed in claim 1, it is characterised in that the data write step is also wrapped Include, when the data production process is to node space to be written write-in data, if the node space to be written with data, directly Connect and write the new original data of data cover.
- 3. the method for interchanging data between multi-process as claimed in claim 1 or 2, it is characterised in that the data production process makes The flag of the node space to be written is accessed with mutual exclusion lock；And the data consuming process accesses the node sky that continues using mutual exclusion lock Between flag.
- 4. the method for interchanging data between multi-process as claimed in claim 3, it is characterised in that the data production process circulation according to It is secondary that data are write in N number of node space and are sent N number of nodal information to data consuming process by domain name socket.
- 5. the method for interchanging data between multi-process as claimed in claim 1 or 2, it is characterised in that the data production process follows Ring writes data in N number of node space and is sent N number of nodal information to data consuming process by domain name socket successively.
- A kind of 6. data exchange system between multi-process, it is characterised in that including：Data consuming process starting module, for starting multiple data consuming process, and it is empty to operating system application shared drive Between；Shared memory space division module, for starting a data production process, the data production process and the data consumption Process defines Data Management Unit, and the form by the shared memory space according to the Data Management Unit using identical rule N number of node space is divided into, wherein N is positive integer；Module is chosen in first node space, for choosing a certain node space as node space to be written；Data write. module, for making the data production process access the flag of the node space to be written, if the node to be written The flag in space is idle condition, then is set in write state, while the data production process is to the node space to be written Write data, by home position be idle condition after the completion of write-in, if the flag of the node space to be written in read states, The node space to be written is abandoned, chooses next node space as the node space to be written；Synchrodata tuple sending module, for repeating to call the Data write. module, the number which is produced According to writing successively in N number of node space, and num node space is often write just by domain name socket by synchrodata tuple The data consuming process is sent to, wherein num is the positive integer for being less than N more than 1；Module is chosen in section point space, for making the data consuming process parse the domain name socket, obtains enabled node sky Between, and according to the enabled node space, a certain node space is chosen as the node space that continues；Data read module, for making the data consuming process access the flag of the node space that continues, the node if this continues The flag in space is then set in read states for idle condition or in read states, while the data consuming process is read simultaneously Data in the node space that continues are handled, by home position are idle condition after the completion of processing, the mark of node space if this continues It is in write state to know position, then abandons the node space that continues, and according to the enabled node space, chooses next node space conduct Continue node space；Replicated blocks, for repeating to call the data read module, until all data consuming process complete the reading of data Take and handle.
- 7. the data exchange system between multi-process as claimed in claim 6, it is characterised in that the Data write. module also wraps Include, when the data production process is to node space to be written write-in data, if the node space to be written with data, directly Connect and write the new original data of data cover.
- 8. the data exchange system between multi-process as claimed in claims 6 or 7, it is characterised in that the data production process makes The flag of the node space to be written is accessed with mutual exclusion lock；And the data consuming process accesses the node sky that continues using mutual exclusion lock Between flag.
- 9. the method for interchanging data between multi-process as claimed in claim 8, it is characterised in that the data production process circulation according to It is secondary that data are write in N number of node space and are sent N number of nodal information to data consuming process by domain name socket.
- 10. the data exchange system between multi-process as claimed in claim 7 or 8, it is characterised in that the data production process follows Ring writes data in N number of node space and is sent N number of nodal information to data consuming process by domain name socket successively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711132105.2A CN107992368A (en) | 2017-11-15 | 2017-11-15 | Method for interchanging data and system between a kind of multi-process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711132105.2A CN107992368A (en) | 2017-11-15 | 2017-11-15 | Method for interchanging data and system between a kind of multi-process |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107992368A true CN107992368A (en) | 2018-05-04 |
Family
ID=62030896
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711132105.2A Pending CN107992368A (en) | 2017-11-15 | 2017-11-15 | Method for interchanging data and system between a kind of multi-process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107992368A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109508241A (en) * | 2018-11-20 | 2019-03-22 | 中国电子科技集团公司第五十四研究所 | A kind of Data Exchange method |
| CN109857698A (en) * | 2019-02-26 | 2019-06-07 | 重庆零壹空间航天科技有限公司 | A kind of rocket flight data processing method, device |
| CN109947575A (en) * | 2019-03-21 | 2019-06-28 | 恒生电子股份有限公司 | Locking, method for releasing and the related system of Read-Write Locks |
| CN112631768A (en) * | 2020-11-23 | 2021-04-09 | 北京思特奇信息技术股份有限公司 | Resource sharing method and system based on asynchronous mechanism |
| CN113132065A (en) * | 2019-12-30 | 2021-07-16 | 西安诺瓦星云科技股份有限公司 | Data communication method, device and system, storage medium and video processing equipment |
| CN115113931A (en) * | 2022-07-22 | 2022-09-27 | 瀚博半导体(上海)有限公司 | Data processing system, method, artificial intelligence chip, electronic device and medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8271336B2 (en) * | 1999-11-22 | 2012-09-18 | Accenture Global Services Gmbh | Increased visibility during order management in a network-based supply chain environment |
| CN103049245A (en) * | 2012-10-25 | 2013-04-17 | 浪潮电子信息产业股份有限公司 | Software performance optimization method based on central processing unit (CPU) multi-core platform |
| CN103218176A (en) * | 2013-04-02 | 2013-07-24 | 中国科学院信息工程研究所 | Data processing method and device |
| CN103414582A (en) * | 2013-07-26 | 2013-11-27 | 北京星网锐捷网络技术有限公司 | Method and device for synchronizing main process and standby process and network device |
| CN103455380A (en) * | 2012-06-05 | 2013-12-18 | 上海斐讯数据通信技术有限公司 | Multi-process communication system and establishment and communication method thereof |
| CN104503734A (en) * | 2014-12-31 | 2015-04-08 | 北京神舟航天软件技术有限公司 | Kahn-based process network program parallel framework extraction technique |
| US20160196131A1 (en) * | 2014-07-07 | 2016-07-07 | Symphony Teleca Corporation | Remote Embedded Device Update Platform Apparatuses, Methods and Systems |
| CN106126359A (en) * | 2016-08-25 | 2016-11-16 | 成都交大光芒科技股份有限公司 | The rapid transmission method of process data sharing |
-
2017
- 2017-11-15 CN CN201711132105.2A patent/CN107992368A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8271336B2 (en) * | 1999-11-22 | 2012-09-18 | Accenture Global Services Gmbh | Increased visibility during order management in a network-based supply chain environment |
| CN103455380A (en) * | 2012-06-05 | 2013-12-18 | 上海斐讯数据通信技术有限公司 | Multi-process communication system and establishment and communication method thereof |
| CN103049245A (en) * | 2012-10-25 | 2013-04-17 | 浪潮电子信息产业股份有限公司 | Software performance optimization method based on central processing unit (CPU) multi-core platform |
| CN103218176A (en) * | 2013-04-02 | 2013-07-24 | 中国科学院信息工程研究所 | Data processing method and device |
| CN103414582A (en) * | 2013-07-26 | 2013-11-27 | 北京星网锐捷网络技术有限公司 | Method and device for synchronizing main process and standby process and network device |
| US20160196131A1 (en) * | 2014-07-07 | 2016-07-07 | Symphony Teleca Corporation | Remote Embedded Device Update Platform Apparatuses, Methods and Systems |
| CN104503734A (en) * | 2014-12-31 | 2015-04-08 | 北京神舟航天软件技术有限公司 | Kahn-based process network program parallel framework extraction technique |
| CN106126359A (en) * | 2016-08-25 | 2016-11-16 | 成都交大光芒科技股份有限公司 | The rapid transmission method of process data sharing |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109508241A (en) * | 2018-11-20 | 2019-03-22 | 中国电子科技集团公司第五十四研究所 | A kind of Data Exchange method |
| CN109508241B (en) * | 2018-11-20 | 2021-04-20 | 中国电子科技集团公司第五十四研究所 | Interprocess data exchange method |
| CN109857698A (en) * | 2019-02-26 | 2019-06-07 | 重庆零壹空间航天科技有限公司 | A kind of rocket flight data processing method, device |
| CN109947575A (en) * | 2019-03-21 | 2019-06-28 | 恒生电子股份有限公司 | Locking, method for releasing and the related system of Read-Write Locks |
| CN113132065A (en) * | 2019-12-30 | 2021-07-16 | 西安诺瓦星云科技股份有限公司 | Data communication method, device and system, storage medium and video processing equipment |
| CN112631768A (en) * | 2020-11-23 | 2021-04-09 | 北京思特奇信息技术股份有限公司 | Resource sharing method and system based on asynchronous mechanism |
| CN115113931A (en) * | 2022-07-22 | 2022-09-27 | 瀚博半导体(上海)有限公司 | Data processing system, method, artificial intelligence chip, electronic device and medium |
| CN115113931B (en) * | 2022-07-22 | 2023-02-14 | 瀚博半导体(上海)有限公司 | Data processing system, method, artificial intelligence chip, electronic device and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107992368A (en) | Method for interchanging data and system between a kind of multi-process | |
| JP4334901B2 (en) | Computer processing system and processing method executed by computer | |
| JP4489399B2 (en) | Data processing method and data processing system in processor | |
| JP3696563B2 (en) | Computer processor and processing device | |
| JP3687990B2 (en) | Memory access mechanism | |
| US8478982B2 (en) | Media access control security management in physical layer | |
| JP4768386B2 (en) | System and apparatus having interface device capable of data communication with external device | |
| JP4499420B2 (en) | Supercharge message exchange device | |
| JP5635521B2 (en) | Method and system for improving latency and reliability of serial port memory communication | |
| US4480304A (en) | Method and means for the retention of locks across system, subsystem, and communication failures in a multiprocessing, multiprogramming, shared data environment | |
| US5418913A (en) | System of two-way communication between processors using a single queue partitioned with pointers and limited overwrite privileges | |
| CN112948318B (en) | RDMA-based data transmission method and device under Linux operating system | |
| CN105183662B (en) | Storage architecture on a kind of Distributed sharing piece of no cache consistency protocols | |
| CN104346317B (en) | Shared resource access method and device | |
| CN111400307B (en) | Persistent hash table access system supporting remote concurrent access | |
| CN104699631A (en) | Storage device and fetching method for multilayered cooperation and sharing in GPDSP (General-Purpose Digital Signal Processor) | |
| JP2003271570A (en) | Data processing method | |
| US20150296018A1 (en) | System for the management of out-of-order traffic in an interconnect network and corresponding method and integrated circuit | |
| CN100458790C (en) | Method and system for access data base | |
| CN113032162B (en) | Multi-process communication method based on shared memory backup mechanism | |
| CN103218337A (en) | SoC (System on Chip) and method for realizing communication between master modules and between slave modules based on wishbone bus | |
| KR100895536B1 (en) | Data transfer mechanism | |
| US7035956B2 (en) | Transmission control circuit, reception control circuit, communications control circuit, and communications control unit | |
| US11385900B2 (en) | Accessing queue data | |
| US6708258B1 (en) | Computer system for eliminating memory read-modify-write operations during packet transfers |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |